profile
viewpoint

gfx-rs/rspirv 358

Rust implementation of SPIR-V module processing functionalities

khyperia/CudaSharp 156

A library to make C# run on CUDA-enabled GPUs

khyperia/Clam 47

A high-quality fractal viewer

khyperia/scopie 33

Astrophotography control software for khyperia's setup

khyperia/Bifrost 3

DCPU mod for KSP

khyperia/khyperia.com 3

Code for khyperia.com - ridiculously simple, because the web is hard and I am not smart

khyperia/pseudogrey 2

Example of pseudogrey colors

khyperia/Astropress 1

An astrophotography stacking program

push eventEmbarkStudios/rust-gpu

khyperia

commit sha 0c65cf3887ba0066cc8d834268c1598dee7b725b

Update .github/CODEOWNERS

view details

push time in 4 days

issue commentEmbarkStudios/rust-gpu

stop relying on spirv-tools

As the error implies, you need to enable "one of these capabilities: Kernel GroupNonUniform SubgroupBallotKHR". You can do that using the API to enable capabilities.

dvc94ch

comment created time in 10 days

issue commentEmbarkStudios/rust-gpu

stop relying on spirv-tools

That's an or between each capability, not an and - you don't need Kernel to use SubgroupSize and friends, just GroupNonUniform. They're available in compute shader SPIR-V and rust-gpu, and pass spirv-val.

dvc94ch

comment created time in 10 days

issue commentEmbarkStudios/rust-gpu

stop relying on spirv-tools

currently the spirv spec says that subgroups are restricted to kernels

can you show where it says that? not seeing where it says that.

dvc94ch

comment created time in 10 days

issue commentEmbarkStudios/rust-gpu

Support Atomics

@dvc94ch Not really - once a path is set down on, it's difficult to re-evaluate and change direction. It takes even more resources to do so, and considering we don't even have enough to do the research in the first place, that's unlikely to happen.

XAMPPRocky

comment created time in 11 days

push eventEmbarkStudios/rust-gpu

khyperia

commit sha 504f56f130b631f0b54088471bbee2c238653563

asm->core::arch::asm

view details

push time in 11 days

PR opened EmbarkStudios/rust-gpu

Reviewers
rustup update

Looks like the clippy update in https://github.com/EmbarkStudios/rust-gpu/issues/796 isn't released yet, but figured I'd bump anyway and get the major work out of the way.

+23 -16

0 comment

7 changed files

pr created time in 11 days

create barnchEmbarkStudios/rust-gpu

branch : rustup-update

created branch time in 11 days

issue closedkhyperia/CudaSharp

Why didn't such a good project continue

Why didn't such a good project continue

closed time in 11 days

phker

push eventEmbarkStudios/rust-gpu

Ashley Hauck

commit sha fe5c7716ed89d8f8ab92374c7ea077e365246b18

Fix ByteAddressableBuffer PassMode::Pair (#837)

view details

push time in 14 days

delete branch EmbarkStudios/rust-gpu

delete branch : bab-pair-fix

delete time in 14 days

push eventEmbarkStudios/rust-gpu

khyperia

commit sha 406433f01ceda205aa54aec7512d478f035a7b8e

Add readme

view details

push time in 14 days

push eventkhyperia/acegikmo-discord-bot

khyperia

commit sha 756e72b241b812921b3e5406d211b5ea778fa594

xtra bonk

view details

push time in 15 days

Pull request review commentEmbarkStudios/rust-gpu

Fix ByteAddressableBuffer PassMode::Pair

+// build-pass++use spirv_std::ByteAddressableBuffer;++pub struct EmptyStruct {}++#[spirv(fragment)]+pub fn load(+    #[spirv(descriptor_set = 0, binding = 0, storage_buffer)] buf: &mut [u32],+    #[spirv(flat)] out: &mut EmptyStruct,+) {+    unsafe {+        let buf = ByteAddressableBuffer::new(buf);+        *out = buf.load(5);

in HLSL, it's documented to be UB to not have a 4-byte-aligned number, but in rust-gpu, I figured hey, UB is gross, let's just document to what HLSL compilers actually do, which is round down to the nearest multiple of 4. So, this is equivalent to passing in 4. (This code is never ran, and it doesn't really matter - all the tests in byte_addressable_buffer are copypasted from each other, and I just chose a random nonzero number without thinking much)

khyperia

comment created time in 17 days

PullRequestReviewEvent

push eventEmbarkStudios/rust-gpu

khyperia

commit sha 208cf7f18e182ed5c90e97383a860da0ffde87f7

whoop

view details

push time in 17 days

create barnchEmbarkStudios/rust-gpu

branch : bab-pair-fix

created branch time in 17 days

push eventkhyperia/acegikmo-discord-bot

khyperia

commit sha 5672282c4bfa7c3930da6ed9b39a8aa12f35edc5

whoopsie, /addgame was broken

view details

push time in 18 days

delete branch EmbarkStudios/rust-gpu

delete branch : update-deps

delete time in 20 days

push eventEmbarkStudios/rust-gpu

Ashley Hauck

commit sha b99fc516e61f6aaf41ad9d58e29edf14d7f3121e

Update deps (#833)

view details

push time in 20 days

PR merged EmbarkStudios/rust-gpu

Reviewers
Update deps

Draft, waiting on https://github.com/EmbarkStudios/ash-molten/pull/63 to be merged (this references the PR hash directly, which will break when the PR is merged)

Merged, now references ash-molten main.

nvm, going to wait for an ash-molten release https://github.com/EmbarkStudios/ash-molten/pull/64

+154 -70

0 comment

5 changed files

khyperia

pr closed time in 20 days

push eventEmbarkStudios/rust-gpu

khyperia

commit sha 51919cce17a72add528d616b7398ae393efdfdb0

Update ash-molten to released version

view details

push time in 20 days

issue commentEmbarkStudios/rust-gpu

Rust Language Support

What about plain function pointers then? AFAIK spirv doesn't support these at all.

Yep, SPIR-V doesn't support them at all right now. With the Slang folks considering pushing through an implementation for dynamic dispatch in SPIR-V, plain function pointers may be able to piggyback on the same infrastructure as well, we'll see (probably more likely than not some hack would make it work). However, we haven't heard any updates in a while (and our unrelated requests for changes/clarifications to the SPIR-V spec have not made much progress at all), so I wouldn't hold your breath.

XAMPPRocky

comment created time in 21 days

push eventEmbarkStudios/rust-gpu

Ashley

commit sha d705a238a1d26aafbb32bc771f8024e0efa7adc6

Add a preserve_bindings opt to the builder and codegen opts (#830)

view details

push time in 21 days

PR merged EmbarkStudios/rust-gpu

Reviewers
Add a preserve_bindings opt to the builder and codegen opts

This resolves https://github.com/EmbarkStudios/rust-gpu/issues/827 and makes reflection a lot easier. I think having it set to true unconditionally is the right move but two alternatives are:

* Setting it to cg_args.spirv_metadata != SpirvMetadata::None. This is pretty unclear behaviour but it does keep things the same in the default case.

  • Adding a new CodegenArgs::opt_preserve_bindings argument. Adds more complexity and is easily missed. Doesn't really 'just work'.
+29 -1

2 comments

3 changed files

expenses

pr closed time in 21 days

PullRequestReviewEvent

PR opened EmbarkStudios/ash-molten

ash-molten v0.12.0+1.1.5

Needed for https://github.com/EmbarkStudios/ark/pull/5707#discussion_r777273694

With the update from ash from 0.33 to 0.35, it's a breaking change, and so needs a minor version bump.

I'll publish this after it's merged.

+1 -1

0 comment

1 changed file

pr created time in 21 days

create barnchEmbarkStudios/ash-molten

branch : v0.12

created branch time in 21 days

more